In the music industry where first week sales has a big impact on overall profitability, this company’s predictions were only 60% accurate thus leaving ‘money on the table’.
The Before State
Models built with internal transactional data were not accurate in the real world.
Created a prediction ensemble utilizing multivariate regression with attributes describing how people ‘engage’ with music in the social space.
The After State
Increase in the accuracy of sales forecast model from 60% to 97% for 133 artists across 10 genres thereby increasing profits and optimizing supply chain.